Label Consistent K-SVD: Learning A Discriminative Dictionary for Recognition

Abstract

A label consistent K-SVD (LC-KSVD) algorithm to learn a discriminative dictionary for sparse coding is presented. In addition to using class labels of training data, we also associate label information with each dictionary item to enforce discriminability in sparse codes during the dictionary learning process. We introduce a new label consistency constraint called discriminative sparse-code error and combine it with reconstruction and classification errors to form a unified objective. The optimal solution is efficiently obtained using K-SVD. Experimental results demonstrate superior performance on face, action, scene, and object recognition tasks.

Examples of Sparse Codes

Each waveform indicates the sum of absolute sparse codes for different testing samples from the same class.

Experimental Results

Feature descriptors for Extended YaleB, AR Face, Caltech101, and Scene15 datasets are used. Spatial pyramid features are constructed and reduced via PCA. Details follow the original paper.

(a) Extended YaleB database

(b) AR database

(c) Caltech101 dataset

(d) Fifteen scene category dataset

Downloads

All materials provided here are for non-commercial research use only.

• Source code1 (LC-KSVD)
• Source code2 (Incremental Dictionary Learning)
• Random face feature extraction
• Random face features for Extended YaleB
• Random face features for AR database
• Spatial pyramid features for Caltech101
• Spatial pyramid features for Scene15

Citation

• Zhuolin Jiang, Zhe Lin, Larry S. Davis. “Learning a Discriminative Dictionary for Sparse Coding via Label Consistent K-SVD”. CVPR 2011. [pdf] [slide] [poster]
• Zhuolin Jiang, Zhe Lin, Larry S. Davis. “Label Consistent K-SVD: Learning A Discriminative Dictionary for Recognition”. IEEE TPAMI, 2013. [pdf]

Questions? Please check frequent questions or contact zhuolinumd@gmail.com.